Use of comorbidity scores for control of confounding in studies using administrative data bases

Background. Comorbidity scores are increasingly used to reduce potential co nfounding in epidemiological research. Our objective was to compare metrica l and practical properties of published comorbidity scores for use in epide miological research with administrative databases. Methods. The literature was searched for studies of the validity of comorbi dity scores as predictors of mortality and health service use, as measured by change in the area under the receiver operating characteristic (ROC) cur ve for dichotomous outcomes, and change in R-2 for continuous outcomes. Results. Six scores were identified, including four versions of the Charlso n Index (CI) which use either the three-digit International Classification of Diseases, Ninth Revision (ICD-9) or the full ICD-9-CM (clinical modifica tion) code, and two versions of the Chronic Disease Score (CDS) which used outpatient pharmacy records. Depending on the population and exposure under study, predictive validities varied between c = 0.64 and c = 0.77 for in-h ospital or 30-day mortality. This is only a slight improvement over age adj ustment. In one study the simple measure 'number of diagnoses' outperformed the CI (c = 0.73 versus c = 0.65). Proprietary scores like Ambulatory Diag nosis Groups and Patient Management Categories do not necessarily perform b etter in predicting mortality. Conclusions. Comorbidity indices are susceptible to a variety of coding err ors. Comorbidity scores, particularly the CDS or D'Hoore's CI based on thre e-digit ICD-9 codes, may be useful in exploratory data analysis. However, r esidual confounding by comorbidity is inevitable, given how these scores ar e derived. How much residual confounding usually remains is something that future studies of comorbidity scores should examine. In any given study, be tter control for confounding can be achieved by deriving study-specific wei ghts, to aggregate comorbidities into groups with similar relative risks of the outcomes of interest.